Aspects of Autonomous Corner Modules As an Enabler for New Vehicle Chassis Solutions

Aspects of Autonomous Corner Modules as an Enabler for New Vehicle Chassis Solutions

Mats Jonasson

Licentiate Thesis

TRITA-AVE 2006:101 ISSN 1651-7660 ISBN 978-91-7178-559-6

Postal address Visiting address Telephone Internet Royal Institute of Technology Teknikringen 8 +46 8 790 6000 www.ave.kth.se Vehicle Dynamics Stockholm Telefax SE-100 44 Stockholm +46 0 790 9290 Typeset using KTH Vehicle Dynamics LATEXstyle

c Mats Jonasson 2007

¡£¢¥¤§¦©¨ ¤

This thesis adopts a novel approach to propelling and controlling the dynamics of a vehicle by using autonomous corner modules (ACM). This configuration is characterised by vehicle controlled functions and distributed actuation and offers active and individual control of steering, camber, propulsion/braking and vertical load. Algorithms which control vehicles with ACMs from a state-space trajectory description are reviewed and further developed. This principle involves force allocation, where forces to each tyre are distributed within their limitations. One force allocation procedure proposed and used is based on a constrained, linear, least-square optimisation, where cost functions are used to favour solutions directed to specific attributes. The ACM configuration reduces tyre force constraints, due to lessen restrictions in wheel kinematics compared to conventional vehicles. Thus, the tyres can generate forces considerably differently, which in turn, enables a new motion pattern. This is used to control vehicle slip and vehicle yaw independently. The ACM shows one important potential; the extraordinary ability to ensure vehicle stability. This is feasible firstly due to closed-loop control of a large number of available actuators and secondly due to better use of adhesion potential. The ability to ensure vehicle stability was demonstrated by creating actuator faults. This thesis also offers an insight in ACM actuators and their interaction, as

a result of the force allocation procedure.

¤¥¢

This research has been accomplished in cooperation with Volvo Car Corpo- ration (VCC) in Göteborg and the division of Vehicle Dynamics at the Royal Institute of Technology (KTH) in Stockholm. The ﬁnancial support was provided by VCC and the Swedish National Energy Agency in the “Energisystem i vägfordon” research programme, which I am most grateful for. I would like to express my gratitude to all people involved, especially my academic supervisor Professor Annika Stensson Trigell at KTH for her support and incessant encouragement during the entire project. I would also like to thank Mr Sigvard Zetterström at VCC for his excellent guidance to the disci- pline of applied chassis technologies and for introducing me into the exciting area of autonomous corner modules (ACM). Furthermore, I would like to thank my co-authors for inspiring partner- ships and for putting my own research into new perspectives, especially Johan Andreasson, PhD student at KTH for guidance in vehicle dynamics and for assisting me with various practical matters. I am also grateful to Dr Oskar Wallmark, now employed by KTH, for fruitful cooperation in merging aspects of control of electrical machines into vehicle dynamics and providing me with useful knowledge of things PhD students must know. I would also like to thank Fredrik Roos, PhD student at KTH for jointly working with active control of vertical wheel loads and for his assistance in understanding mechatronics. I am indebted to Mr Alf Söderström at VCC for his insights into vehicle handling. I would also like to thank all members of the steering committee, who have not been acknowledged before, for their valuable guidance; Professor Mats Alaküla at LTH, Assistant Professor Urban Kristiansson at VCC and Mr Johan Wedlin at Volvo AB. I would also like to thank Indira Antonsson at VCC for proofreading this thesis and for providing English language advice.

Finally, I’m very proud and fortunate to be married to my wonderful wife Cathrin, who has given me support, care and understanding throughout. I am blessed by the presence of my three children; Frida, Simon and Soﬁa, my lights of inspiration. I thank you for your constant support. Thank you all for your assistance, which has undoubtedly contributed to this licentiate thesis.

Mats Jonasson Göteborg, February 2007

! £¨! !"¦©¢

Paper A Jonasson, M., Zetterström, S. and Trigell, A. S., ‘Autonomous corner modules as an enabler for new vehicle chassis solutions’, Proceedings of the FISITA World Automotive Conference, Yokohama, Japan, October 2006.

Paper B Jonasson, M. and Wallmark, O., ‘Stability of an electric vehicle with permanent magnet in-wheel motors during electrical faults’, Proceedings of the 22nd International Battery, Hybrid and Fuel Cell Electric Vehicle Symposium and Exposition, Yokohama, Japan, October 2006.

Paper C Jonasson, M. and Wallmark, O., ‘Control of electric vehicles with autonomous corner modules: implementation aspects and fault handling’, submitted for publication, 2006.

Paper D Jonasson, M. and Andreasson, J., ‘Exploiting autonomous corner modules to resolve force constraints in the tyre contact patch’, submitted for publication,

2007. #%$'&)(+*-,/.10)(2,3$'&546$87¥,9&5:',<;5,3:©05=8>?=@0)(+A5$8*B4

Mats Jonasson, PhD student, is the ﬁrst author of all appended papers. In [paper A], he undertook all writing, while Mr Sigvard Zetterström and Professor Annika Stensson Trigell contributed with the problem formulation and useful background material. In addition, they contributed with valuable comments

and took part in reviewing the paper. C The contributions of the individual authors in [paper B and C], as concerns modelling and writing the paper, are equal. In these papers, Dr Oskar Wall- mark has described the electrical parts used and their behaviour during electrical faults. In addition, he has performed measurements on electrical machines and simulations of power electronics. Mats Jonasson has contributed to the method of tyre force allocation adopted and the tyre models used. In [paper D], the contribution from both authors are equal. Mats Jonasson has prepared all models and simulations and the paper has been jointly writ- ten. Johan Andreasson, PhD student, has given essential inputs to the approach adopted. One of them is related to the interpretation of vertical tyre force allocation and the subsequent evaluation of adhesion potential applied.

E¤¥¤¥¢

FHG+IKJ-LNM+OPJ Q

FRO2S2T+UWVYX Z2[]\^Z`_aZ`T2JKI QbQbQ

FHc]c]Z`T`[^Z2[§c+M^c]Z`LNI d

UfT2JKZ`T2JKI dgQbQ

h h

jilknmpo+q+rtsuknv o2i wtx

...... w

9 j

wtx ......

5KK H-3W-P{P

wtx ......

wtx `x

......

wtx `x ){P uK W Rb?lP {-9 B B}n9 }©P } K

. . . . .

wtx Ku{-3}H{PP-3t{-}

...... ¡

wtx x

...... ¤

wtx x ¦¥§9 -} P ?¨fP } ?}BW B3W

...... ©

wtx x ª§}n{B W « -'}Bl9 -N¬2uK -Rª)®¯b«P -{Bb¨9l3}BK°w-±

²´³ hh

rWµgµ·¶Pm9¸Hot¹®¶ºWº2»litqW»q¥ºt¶º2»lmp¼

½´³ hP½

sv »lilknv ¾`sHs-o2ilknmNvÀ¿Wrlknv o2it¼

hÆ

ÁÂ)o2itsÃÀrtq+vÀitÄamp»lµ·¶Pm£Å¼

ÆÈÇ hPÊ

»s-o2µgµ·»litqW¶Pknv o2it¼¹potm)¹NrlknrWmp»HÉotm£Å

Ë5ZÌNZ`LNZ`T`OlZWI ÍfÎ

Â)o2ilk»lilk¼

ÏÑÐ§ÒÔÓ°Õ?Ö6×ÙØ Ú)¤§¦©ÛEÜ ¤ÞÝE

This chapter provides the background for this research and introduces drivers to advanced chassis technology where functions of vehicle dynamics are software controlled. Contributions of design proposals exploiting this opportunity are discussed. Finally, a brief description of the research approach adopted is

presented.

ßàbß á%=®âäã|å©*B$'0&5:

This research started at Volvo Car Corporation (VCC) in 2004, after a long pe- riod of promising ACM design proposals from the inventor Mr Sigvard Zetter- ström at VCC. The reason for the following research has been to prepare the vehicle for the transition into a new type of chassis, partly driven by new propulsion technologies and energy buffers. Keeping the approaching shortage of fos- sil fuels and the environmental challenges in mind, there is a need to understand chassis implications in the use of alternative propulsions than those offered by the combustion engine. New energy converters and buffers set new conditions for the chassis development in its entirety. In addition, there has been an interest to understand the implications of vehicle conﬁgurations, where each corner is able to act independently without any mechanical connections between the wheels and between wheels and power source. At the same time, Professor Annika Stensson Trigell and Johan Andreas- son, PhD student, at KTH were in the process of investigating a new research approach of controlling vehicles involving methods of tyre force allocation.

The thoughts behind these ideas came from knowledge of ﬂight control. Thus, h

² æfçèé-êNë-ìäí îðï-êbì/ñtòtóônêbõ ñlï

KTH started a joint research project, supported by the Swedish National Energy

Agency, designating ACM as the target for research.

ßàpö ÷Þ*-,<;)ø5*B4

The increasingly sophisticated demands on chassis technologies are driving the development process of vehicles toward new and refined chassis solutions for even more modern vehicle platforms. Since nearly all chassis systems include mechanical components, a great deal of replacements are involved. In addition, modifications of chassis components in existing vehicles are difficult to carry through since the mechanical components can rarely be tuned to perfection. As a result, there has been an increasing interest in new chassis solutions, as an alternative to the state-of-the-art options, where attributes of vehicle dynamics are determined by software implementations. There has been extensive research into the use of active chassis components to enable new vehicle dynamic functions in conventional vehicles. However, vehicles that allow each tyre to generate forces in perpendicular directions and also allow all tyres to be fully independently controlled, have not gained equal attention. These types of vehicles, with “freely controlled” tyre forces, enable a new and broader range of possible vehicle responses that cannot be performed in conventional vehicles. One example of this is a new motion pattern for vehicles for low and high speeds respectively. Since conventional vehicles are front-steered, there are limitations to the motions that would be feasible if wheel kinematic constraints are alleviated. If the rear wheels were allowed to steer, parking manoeuvres would be enhanced at low speeds. Also, the vehicle slip would be better assigned for different driving scenarios at high speed. If tyre forces were allowed to be freely controlled and also in a closed-loop system, even more new opportunities of vehicle dynamics could be enabled. One of them is vehicle robustness to disturbances, which may expose the vehicle to discomfort or to the risk of stability loss. Disturbances originate from external vehicle excitations, such as side wind, or internal sources, which may be caused by faults of vital functions. In such circumstances, conventional vehicles require driver corrections of deviations from the desired motion. However, the correction needed is restricted to the driver’s capability of handling a vehicle and also to the few actuators that can be used for recovery. Alternatively, in vehicles where tyre forces are allowed to be freely controlled in a closed-loop, necessary corrections could be rapidly executed by all available actuators from vehicle control commands instead. This would be supported by the driver who acts relatively slowly, compared with the actuators, in an outer loop. Thus, this

ëuônêbõ ñlïHíPú û üäëBýþõ ÿtïHéuì/ñlélñýbè¡ ý ½ ù

new type of vehicle conﬁguration would be capable of relieving pressure from the driver regarding his/her control task even during conditions where disturbances are present. As a result, vehicle comfort and stability could be better ensured, which clearly can be used to support the implementation of driver aid systems. The conventional vehicle is characterised by signiﬁcantly high limitations on tyre force generation, which in turn depends on limitations of wheel geometry and kinematics. One way to partially accomplish this situation is to add active chassis components to conventional vehicles in order to resolve tyre force restrictions. Consequently, vehicle design becomes more technically complex due to the fact that the conventional chassis has many constraints. From this discussion it is evident that chassis designs, where functions are mainly software implemented and tyre forces are allowed to be freely con-

trolled, are of considerable interest.

ßà£¢ ÷ øä42,3å©&¥¤5*B$¦¤'$@4=8>34

Bearing in mind the discussion above, many inventions involving this class of chassis conﬁguration have been developed. One contribution is the autonomous corner module (ACM), which was invented at VCC in 1998 [1]. The name autonomous indicates that wheel forces and kinematics are individually controlled, supporting a common task. The design proposal, as illustrated in Figure 1.1 is designed to offer:

(i) A chassis design that could be reused in the development process for new vehicle platforms.

(ii) Functions to be regulated by software.

(iii) Individual control of each wheel; propulsion/braking, steering/camber and vertical wheel load.

This is solved by the use of a hub in-wheel motor combined with actuators for steering/camber and vertical wheel load [2, 3]. Another contribution is the eCorner [4, 5] as illustrated in Figure 1.2a, which is an integrated solution with an in-wheel motor and actuators on each wheel for steering and vertical wheel load combined with a wedge brake. How- ever, wheel camber can not be controlled in this concept. In addition, the concept car HY-LIGHT r must be mentioned, which involves a solution where in-wheel motors are combined with the individual control of the wheel suspension [6]. The vertical alignment of the wheel, provided

æfçèé-êNë-ìäí îðï-êbì/ñtòtóônêbõ ñlï Á

Figure 1.1: (a) The ACM disclosure from patent application [1] and (b) The ACM concept further developed in cooperation with MAGNA STEYR (illustration reproduced with permission of MAGNA STEYR Fahrzeugtechnik AG & Co KG, Engineering, Advanced Development Chassis) by an electrical actuator, as seen in Figure 1.2b, is supported by a transverse control arm, which can be used to roll the car body as desired (not illustrated in Figure 1.2b). Wheel camber cannot be controlled in this concept and steering

is an ordinary rack-and-pinion.

Î¨§ © §Î UfT`OlZ`c2JHO +M2LNM+OPJKZ`L IBMWJ UfT

In order to characterise and compare the design proposals’ abilities to move and control the wheel along different axes and directions, these concepts are presented in Table 1.1. Here, a conventional premium vehicle is added to Ta- ble 1.1 as a comparison. This vehicle is assumed to be four wheel driven and provided with a McPherson front strut and multilink rear axle. For some concepts, motion is a result from actuators performing the main mechanical work in the close vicinity of the wheel. Here, this attribute is denoted distributed actuation. In some cases the degree of freedom of the wheel is denoted as vehicle controlled (denoted as V in Table 1.1), which indicates that the actuators’ control inputs are composed of a combination of signals from other co-existing vehicle functions. One example to consider is vehicle

ëuônêbõ ñlïHíPú û üäëBýþõ ÿtïHéuì/ñlélñýbè¡ ý Æ ù

Figure 1.2: (a) The Siemens VDO eCorner: (1) wheel rim, (2) in-wheel motor, (3) electronic wedge brake, (4) active suspension, (5) electronic steering [5] and (b) Michelin’s wheel unit HY-LIGHTr (illustrations reproduced with permission of Siemens VDO and Michelin respectively) controlled steering, where signals from the vehicle and its environment can be combined with the steering wheel angle to create the steering actuator control input. Thus, the driver is assisted in his/her control task to follow the anticipated vehicle response. On the contrary, driver controlled steering (denoted as D in Table 1.1) is characterised by the use of steering wheel signal only to determine the control input. The use of vehicle controlled functions embraces a variety of control structures, which has been classified in depth for vehicles in [7]. Table 1.1 even shows motion restrictions between wheels, which are present for some concepts. Wheel camber is predetermined and bound by the suspension kinematics (denoted as B in Table 1.1) for the McPherson and the multilink suspensions. Thus, the resulting wheel camber depends on the wheel travel only. Suspen- sions, where camber is invariable for the entire wheel travel range with respect to car body coordinates, is denoted as invariable in Table 1.1. However, under the influence of lateral tyre forces, all these concepts influence the camber in different ways due to the suspension elastokinematics. This effect is normally not desired when handling the vehicle. However, if camber is allowed to be vehicle controlled, this effect can be suppressed and camber can be aligned to the performance needed.

æfçèé-êNë-ìäí îðï-êbì/ñtòtóônêbõ ñlï

Table 1.1: Degrees of freedom for front and rear wheels, V=Vehicle controlled, D=Driver controlled, I=Invariable with respect to car body coordinates, B=Bound by wheel kinematics ACM eCorner HY-LIGHT r Conventional front/rear front/rear front/rear front/rear Wheel spin V/V V/V V/V V1,2/V1,2 Steering V/V V/V D2/B D2/B Camber V/V I/I I/I B/B Vertical dynamics V/V V/V V/V V3/V3 1The left–right distribution of wheel torques can be changed by braking. 2Restrictions between wheels. Fixed relation between left and right steering angle. 3Semi-active damping instead of active damping and levelling.

The concepts’ ability of controlling vertical loads in a force–velocity graph representation is often referred to as active damping and semi-active damping respectively [8]. The former represent a class of dampers, that requires energy to be supplied directly to the actuator of the targeted actuation. Thus, the damper is able to generate forces in all four quadrants. In contrast, semi-active dampers do not require power supply, except for energy needed to alter the desired damper rates. Normally, this damper dissipates energy by converting it to heat. The semi-active damper is due to the non-demanding energy consumption, restricted to operate in two quadrants in the force–velocity graph. Passive dampers are restricted to operate along predeﬁned curves in two quadrants. From the discussion above, it is evident that all these novel design proposals demonstrate three things which are typical for this class of chassis conﬁg- uration; vehicle controlled functions, few motion restrictions and distributed

actuation.

Î¨§ © §<Í @LNM^T+I J UfT JKU6_aU§[^X M2L IBZ2[ Z+X Z2OPJ-LÀOWZ ÀOX ZWI

The transition from conventional vehicles to vehicles with distributed actuation, as illustrated in Figure 1.3, implies functions to be physically implemented to the vehicle corners. This, in turn, enables the development of a modularised chassis conﬁguration, where the number of chassis components, from a car manufacturer’s perspective, can be reduced. Vehicle controlled functions facilitate transfer functions between driver and the actuators in question, to be more freely adopted. In addition, mechanical connections can then be replaced by by-wire technology. Figure 1.3 illustrates the ACM vehicle, where all chassis functions are by-wire controlled. This ve-

ëuônêbõ ñlïHíPú +ëBý£ëuèBì/ôuç«èéPéuì/ñè-ôuç Ê ù

hicle has the ability to change transfer functions between the driver instruments to the vehicle controller, e.g. steering characteristics. In general, this is not possible for conventional vehicles, where these transfer functions are ﬁxed and implemented by mechanical links.

Figure 1.3: ACM vehicle with vehicle controlled functions and distributed actuation (illustration by Sigvard Zetterström)

These design proposals also indicate a trend towards electric actuation. Nevertheless, functions can be developed using other power sources than the electric counterparts, e.g. hydraulic power. However, pure electric concepts are flexible for different vehicle configurations, such as; hybrid electric vehicles, electric vehicles and fuel-cell vehicles. Generally the electric concepts also provide substantially high energy efficiency. In addition, the access to high voltage power sources in vehicles facilitate the use of high power electric

actuators.

ßà ·øä4øä=)*BâäA = ¤!¤5*B$@=®âäA

It is a well-known fact that vehicle motion is a result of tyre forces generated in the contact patch between tyre and ground. However, the constitution of vehicles are characterised by engines and motors packaged at a distance from the tyre contact patch. Given this situation, the research question is as follows:

If tyre forces are allowed to be controlled more freely and chassis functions are distributed, what possibilities and consequences of vehicle dynamics can be obtained?

" æfçèé-êNë-ìäí îðï-êbì/ñtòtóônêbõ ñlï

This research originates from the tyre forces and the ability to control them, involving limitations of tyre forces, from the tyre itself and indirectly by actuators, which are used to generate the tyre force under consideration. This research seeks to investigate how the range of vehicle dynamic functions are related to tyre force generation, and in turn, bring more knowledge into different chassis concepts supporting tyre force generation. These chassis concepts involve components to generate propulsion torque, lateral tyre forces and verti-

cal load independently at each wheel.

Î¨§$#%§Î &2U^L£OlZ OlUfT+IKJ-LNM

To be able to freely generate tyre forces, limitations of traditional wheel kinematics must be resolved. Consider the following; conventional vehicles cannot steer the rear wheels. Consequently, lateral tyre forces cannot be generated there, except for forces generated by the vehicle state (when rear axle slip is present). If tyres were exploited to the utmost, forces could be generated in directions, where the boundaries, which correspond to the tyre force limitations, are depicted by surfaces stretched in the x-y-z space as illustrated in Figure 1.4.

Figure 1.4: Tyre force constraints

As seen, these surfaces resemble cones, where the length of them are related to vertical tyre load and friction coefﬁcient. The cross-section corresponds to

ëuônêbõ ñlïHíPú +ëBý£ëuèBì/ôuç«èéPéuì/ñè-ôuç ' ù

the well-known friction ellipse [9]. Ideally, the vehicle is able to utilise the tyre to the utmost limits. This means, if tyre forces are represented as vectors growing from the cone’s tip, these forces are allowed to be generated inside the cones. However, limitations in actuators may set constraints on the tyre resources of force generation, and in turn, the cones become restricted [10]. One example of these restrictions is in-wheel motors with limited peak torque, which end up with curtailed cones along the longitudinal axis, due to limitations in possible longitudinal slip and

hence, the maximum longitudinal force.

Î¨§$#%§<Í (L

From interactions via driver instruments and controls, the desired vehicle motion can be computed into a state-space trajectory, which span over a number of degrees of freedom [11]. This trajectory can be conceived to be an effect arising from global forces acting on the vehicle’s centre of gravity. These forces are then distributed, via the force allocation procedure (see Figure 1.5) to each tyre [12]. This procedure is complex due to inﬁnite sets of tyre forces, which can end up in the same global forces. Thus, due to the mathematically under-determined nature of the force allocation problem formulation, the solver methodology needs to be thoroughly scrutinised. A suggested solver methodology in this thesis is to use constrained, linear, least-square optimisation, similar to the methodology used in [13]. As previ- ously discussed, the tyre and actuator are not limitless. Therefore, these limitations need to be incorporated into the force allocation problem formulation. Various optimisation methods for solving similar force allocation problems for ﬂight control have been evaluated in [14, 15].

Constraints Reference andcost trajectory Actual functions trajectory

Actuator Path Force Vehicle andtyre control allocation dynamics dynamics

Referencesignals Actualsignals

Figure 1.5: Principle of vehicle control

h+* æfçèé-êNë-ìäí îðï-êbì/ñtòtóônêbõ ñlï

In order to drive the force allocation procedure towards a desired solution, additional cost functions are added to reformulate the problem. Thus, solutions that contribute to the designated cost functions are prioritised above other solutions. This can be used to set a favourable vehicle response. In this thesis, cost functions are used to favour small force deviations from nominal levels and better utilise adhesion potential in the tyre contact patch. The latter is used to ensure the stability margin of the vehicle, which has been elaborated in [16, 17]. Depending on the adopted cost functions, an exact solution of the force allocation formulation may not exist. The path controller studied [paper B], as illustrated in Figure 1.5, is designed to minimise deviations of the desired trajectory by changing its outputs of global force references. Controllers adopted for similar tasks have been in-

vestigated in [13, 18].

e10

Î¨§$#%§ © FROPJ+MWJ UfT,

Actuators in the ACM are used to generate the reference forces; the in-wheel motor generates longitudinal tyre forces, steering actuators generate lateral tyre forces and ﬁnally one actuator controls the vertical tyre forces. These actuators are controlled by compiling the reference tyre forces into tyre slip angles and vertical tyre force needed. This is accomplished by inverse tyre models, where the tyre slip angles are separated from one another and assigned to each actuator, as seen in Figure 1.6.

Wheeli (i=1,2,3,4)

Longitudinaltyre forcereference In-wheelmotorand frictionbrakedynamics Wheeltorque Inverse Lateraltyreforce tyremodels reference Wheel Steeringactuator geometry dynamics Steering Verticaltyreforce reference Verticalload Verticalload dynamics

Frictioncoefficientin thetyrecontactpatch

Figure 1.6: Force assignment to ACM actuators

ÏÑÐ§ÒÔÓ°Õ?Ö6×32

Ü ¨£¦!5 76 ¨! ! £¨! !"¦©¢

= ¤'ø5*:9<;/=>9Þ0)($'&5$@? $'054 âf$8*&5ø5*A? $©:©0@>3øä4 =®4 =@&¯ø5&5=8.>3ø5*

7K$8* &5ø B ;)ø5A@,3â|>3øâäA5=®442,3464$1>90)(2,3$'&54¡C

This paper describes and reviews associated findings. Research findings indicate that the ACM is capable of covering a extensive range of “stand-alone” functions, which have been used for conventional vehicles to achieve benefits in vehicle dynamics. Associated findings suggest methods for chassis control, where tyre forces can be allocated from a vehicle trajectory description. In addition, these indicate that the ACM introduces new opportunities and shows

itself to be a promising enabler for vehicle dynamic functions.

= ¤'ø5* áD;E= F'(=8.,/>/,<(G $87 =@& ø®>3øäâ](+*-,3â ;)ø5A@,3â|>3øHB6,<(+AI¤'ø5*¡? =%J

&5ø5&)(K? =®å©&5ø§(¥,9&LJ¡B A5øäø®>E? $5($8*B4 :©0*-,9&5å ø®>3øäâ](+*-,3âf=8>)7K=@0@><(4¡C

This paper presents an analysis of the stability of an electric vehicle equipped with in-wheel motors and individual steering actuators. The vehicle stability has been evaluated by simulations via an injection of an electrical fault, arising from a 30 kW electric permanent-magnet synchronous-machine. It is shown that the electrical fault risks a major decline in vehicle stability if the correct counteraction is not quickly undertaken. However, by introducing a path controller combined with an unsophisticated procedure of tyre force allocation, stability can be maintained. This inherent capacity to handle faults is attractive,

especially since no additional fault handling strategy or hardware is needed. hh

hP² æfçèé-êNë-ì M ó+N1NRèBìPO'ñQ§èéPéPëïòëuòHéèéPë-ìý

= ¤'ø5*?#D;=2#%$'&)(+*B$1>^$87'ø®>3øäâ](+*-,3â ;)ø5A@,3â|>3øä4:B6,<(+A§=@0)($'&5$@? $'054

âf$8*&5ø5*R? $©:©0@>3øä4¨;Ù,?S¤>3øE? ø5&)(=ä(2,3$'&Ù=®4T¤'øäâ](4=@&5:Ù7K=@0@><(

A5=@&5:'>/,9&5åUC

In this paper, the ACM is studied adopting the procedures of tyre force allocation by optimisation, where a linear description of tyre and actuator constraints is formulated. In order to evaluate the proposed vehicle control principle [paper B], the ACM vehicle is exposed to realistic fault conditions. If these conditions occur during extreme driving scenarios, the path controller can hardly maintain stability. However, if the constraints in the optimisation procedure used for tyre force allocation are adapted to the speciﬁc fault, stability is secured. The paper also demonstrates how limited computational capacity used in the optimisation solver can result in unwanted interactions between the individual actuators and

thus, indirectly affect vehicle stability.

= ¤'ø5*Þ÷<;@=¨VXWX¤>3$1,<(2,9&5å =@0)($'&5$@? $'054§âf$8*&5ø5*D? $©:©0@>3øä4°($

*Bøä4$1><;)ø 7K$8*Bâfø âf$'&54l(+*B=8,9&)(4Ô,9&(+A5ø (G8*Bø âf$'&)(=®â](Y¤8=ä(âäA C

In this paper, a vehicle exploiting ACMs has been forced to follow a trajectory, identical to a conventional front-steered vehicle trajectory under an evasive manoeuvre. Tyre force allocation by optimisation has been adopted, which also involves vertical tyre forces, along with tyre force constraints and cost functions to favour a desired solution. This has been evaluated as open-loop, where tyre forces needed are identiﬁed and they are shown to be allocated in a different manner than in conventional front-steered vehicles. A suggested approach for control of steering actuators is presented, where the actuator limitation is related to the lateral force possible. Finally, the force allocation strategy involves the ability to control vehicle slip independently from vehicle yaw rate, which is used to increase adhesion potential.

ÏÑÐ§ÒÔÓ°Õ?Ö6×3Z

Ý¤ÞÝ¡[ E¤§¦HÝ+¡!Ü¤ÞÝE ¢

The main scientiﬁc contributions in this thesis can be summarised as follows:

1. The ACM capability and its possible uses related to associated research ﬁndings [paper A].

2. The closed-loop control principle of ACM using a simple procedure of force allocation [paper B], is shown to have an in-built robustness to faults. This inherent capacity can ensure vehicle stability with no need of additional fault handling strategy or hardware.

3. The control of ACMs particulary in on-board implementations are described in [paper C]. Force constraints are linearised and sample frequency is considered to reduce computational effort. Sample frequency is related to actuator interaction and vehicle stability.

4. Feedback of actuator constraints into the force allocation procedure is of high importance even when faults occur. Without adaptation of actuator constraints vehicle stability is threatened, especially during extreme driving conditions [paper C].

5. It is demonstrated that the ACM is capable of imitating motion characteristics of a conventional vehicle trajectory, but with better use of adhesion potential. By reducing vehicle slip, the ACM even demonstrates an im-

provement in adhesion potential [paper D]. hP½

h æfçèé-êNë-ì§û ôBõ ëï-êbõ \ô'ôKñlï-êbì9õ ]Pó-êbõ ñlï-ý

Á ù

ÏÑÐ§ÒÔÓ°Õ?Ö6×_^

This thesis presents a novel class of vehicles with the ability to control tyre forces more freely than in conventional vehicles. One contribution to this technology is the autonomous corner modules, which is adopted and described in this thesis. This specific vehicle configuration is, because of all functions implemented in software, fully dependent on appropriate control. A control principle of ACM vehicles is developed and evaluated involving a procedure of tyre force allocation and control of the ACM actuators. The range of possible motions for conventional vehicles is limited due to restrictions of wheel kinematics, and in turn, constraints in tyre forces. The constitution of the ACM and the adopted control principle is shown to be able to resolve tyre force constraints. Hence, an extensive range of vehicle motion is enabled, which can be used to change motion characteristics. One example of this is for cornering manoeuvres. Here, the ACM vehicle demonstrates its capacity to control vehicle slip and yaw independently from one another. This is attractive, since driver perception can be dynamically changed. In addition, these quantities can be differently mixed in order to assign specific motion characteristics during hazardous events. Under extreme conditions, the ACM vehicle is shown to be capable of diagonal movements, it rapidly executes a lateral manoeuvre and still maintains a modest yaw movement. The tyre force allocation procedure is flexible and can be used to utilise adhesion potential better than in conventional vehicles. Thus, the margins of the tyre force constraints are kept equal and vehicle stability can be better maintained. It is also shown that the ACM vehicle has an extraordinarily inherent capacity to handle faults. This capacity is useful since electrical faults are com-

monly considered to be a well-know problem for this type of vehicle conﬁgura- hÆ

h+ æfçèé-êNë-ì æ]ñlïô` óòlõ ïÿìë¡NRèBì£aKý

tion. One conceivable risk is the entry of electrical machines with permanent- magnets, which can endanger vehicle stability if the control of them fails. How- ever, the results presented in this thesis show that the ACM control principle along with adaptation of the constraints for the specific fault, cancel out these risks. Obviously, the development of ACM vehicles addresses new challenges. One of them is related to the driver interface and the anticipated response of vehicle and driver controllers. Another challenge is also to offer a vehicle with sufficient degree of dependability. In addition, the development of electric actuators will meet difficulties due to the demands on high force and compact design. Nevertheless, the results presented in this thesis show that the ACM solution offers new opportunities and enables new vehicle dynamic functions.

ÏÑÐ§ÒÔÓ°Õ?Ö6×cb

!¦·

9aâ](+05=ä($8*B46=@&5: (G8*Bøä4

• The tyres considered in this work has been modelled as ordinary car tyres. However, tyres for conventional vehicles are developed for best performance involving a variety of compromises for attributes, such as; low rolling resistance, low tyre wear and sufﬁcient grip. This compro- mise is not based on a tyre where wheel kinematics is expanded, especially for dynamical alignments of wheel camber. Given this situation, there is a need to investigate if a novel tyre design can offer any conceivable options.

• It is of considerable interest to reduce power consumption of ACM actuators. One area that could be studied is; the process of steering the wheel by moving steering actuators non-symmetrically combined with camber

variations [19].

$5(2,3$'& âf$'&)(+*B$1> f

• In this thesis, robustness to faults has been evaluated in-depth. However, there is a considerable interest to further investigate the capacity of the

ACM vehicle to handle disturbances in a wider perspective, e.g. in the hPÊ

case of a puncture or exposure from side wind.

• Involve predictive control methods, where the vehicle states can be pre- dicted. These methods beneﬁt from adjusting the force constraints in order to optimise the vehicle response in advance.

• The tyre force models applied in this thesis are not valid for low speed (approximately below 5 km/h). It is of considerable interest to further develop and evaluate the control principle using appropriate tyre models for low speed manoeuvres.

• The control task involves dead times. These originate from delays in ACM actuators, such as friction brakes, and also deterministic delay in data bus. Since these dead times can hardly be tackled by the proposed control principle, further improvement is suggested.

• The ACM ability to control vehicle slip and vehicle yaw individually is most interesting, since the perception of cornering can be addressed to the speciﬁc brand. This ability enables new functions of active vehicle safety. Therefore, there is a considerable interest to develop an inherent control method for this matter.

• The driver interface is not considered in this thesis. However, by merging driver models into the ACM models, vehicle response could be evaluated

under more realistic conditions.

• The sensor signals used, such as vehicle state and friction coefﬁcient, have been assumed to be ideal without any errors. Realistic implementations comprise sensors and state observers, where signals need not be accurately measured and estimated. Hence, it is highly interesting to evaluate the vehicle response to errors for these signals, thus the importance of the different estimates can be rated.

h+'

• The ACM is characterised by high degrees of freedom as concerns wheel kinematics. If the number of degrees of freedom, for some reason needs to be reduced, it is would be of particular interest to categorise the adopted restrictions to the implications of vehicle dynamics.

• In this thesis, vertical dynamics involving the sprung and unsprung masses are intentionally ignored. It is worth considering the effect of these dynamics for vehicle response and also developing control methods, where the state of the car body and the unsprung masses can be controlled.

• There is a deﬁnite need to validate the ACM vehicle, by implementing computer models in physical demonstrators.

d I6©"¦© ¢

[1] Zetterström, S., ‘Vehicle wheel suspension arrangement’, patent no. EP1144212, 1998.

[2] Zetterström, S., ‘Electromechanical steering, suspension, drive and brake modules’, Proceedings of the 56th IEEE Vehicular Technology Confer- ence, Vol. 3, pp. 1856–1863, Vancouver, Canada, September 2002.

[3] ‘Volvo styr med alla hjulen’, Ny Teknik (in Swedish), February 22, 2006. Also availible on the Internet: http://www.nyteknik.se.

[4] Automotive engineer, Vol 31, No 8, 2006.

[5] Siemens VDO, ‘Car motors will disappear into the wheels: Siemens VDO starts eCorner development’, availible on the Internet: http://www.siemensvdo.com, December 2006.

[6] Dietrich, P., et. al., ‘Concept, design and ﬁrst result of the light-weight- fuell-cell vehicle Hy-light’, Proceedings of the 21nd International Bat- tery, Hybrid and Fuel Cell Electric Vehicle Symposium and Exposition, Monaco, April 2005.

[7] Andreasson, J., Knobel, C. and Bünte, T., ‘On road vehicle motion control–striving towards synergy’, Proceedings of 8th International Sym- posium on Advanced Vehicle Control, AVEC’06, Taipei, Taiwan, 2006.

[8] Lizell, M., ‘Dynamic levelling for ground vehicles, a low power active suspension system with adaptive control’, PhD dissertation thesis, TRITA-MAE-1990-9, ISSN 0282-0048, Department of Machine Ele- ments, KTH, Sweden, 1990.

[9] Pacejka, H. B., Tyre and vehicle dynamics, Butterworth-Heinemann, Ox-

ford, United Kingdom, 2002. ²+h

²² Ç »¹p»lmp»lits-»¼

[10] Andreasson, J., Laine, L., Fredriksson, J.,‘Evaluation of a generic vehicle motion control architecture’, Proceedings of the FISITA 2004 World Automotive Congress, Barcelona, Spain, May 2004.

[11] Sharp, R. S, Casanova, D. and Symonds, P., ‘A mathematical model for driver steering control, with design, tuning and performance results’, Journal of Vehicle System Dynamics, Vol. 33, pp. 289–326, 2000.

[12] Knobel, C., Pruckner, A. and Bünte, T., ‘Optimized force allocation. A general approach to control and to investigate the motion of over-actuated vehicles’, Proceedings of the 4th IFAC Symposium on Mechatronics, Hei- delberg, Germany, September 2006.

[13] Andreasson, J. and Bünte, T., ‘Global chassis control based on inverse vehicle dynamics models’, to be published in Journal of Vehicle System Dynamics, 2007.

[14] Bodson, M., ‘Evaluation of optimization methods for control allocation’, Journal of Guidance, Control and dynamics, Vol. 25, No 4, 2002.

[15] Härkegård, O.,‘Backstepping and control allocation with application to ﬂight control’, PhD dissertation thesis, ISBN 91-7373-647-3, Department of Electrical Engineering, Linköping University, Sweden, 2003.

[16] Orend, R., ‘Vehicle motion feedforward control with minimum utilisation of the friction potential at all four tyres’, Journal of Vdi Berichte, Vol. 1828, pp. 475-484, 2004.

[17] He, P., Improvement of EV maneuverability and safety by disturbance observer based dynamic force distribution, Proceedings of the 22nd Inter- national Battery, Hybrid and Fuel Cell Electric Vehicle Symposium and Exposition, Yokohama, Japan, October 2006.

[18] Fredriksson, J., Andreasson, J. and Laine, L., ‘Wheel force distribution for improved handling in a hybrid electric vehicle using nonlinear control’, Proceedings of the 43rd IEEE Conference on Decision and Control, Nassau, Bahamas, Vol. 4, pp. 4081–4086, December 2004.

[19] Pacejca, H. B, ‘Spin: camber and turning’, Journal of Vehicle System Dy- namics, Vol. 43, Supplement, pp. 3–17, 2005. Jonasson, M., Zetterström, S. and Trigell, A. S., ‘Autonomous corner modules as an enabler for new vehicle chassis solutions’, Proceed- ings of the FISITA World Automotive Conference, Yokohama, Japan, Paper A

October 2006. ²½

² Á Jonasson, M. and Wallmark, O., ‘Stability of an electric vehicle with permanent magnet in-wheel motors during electrical faults’, Pro- ceedings of the 22nd International Battery, Hybrid and Fuel Cell PaperB Electric Vehicle Symposium and Exposition, Yokohama, Japan, Oc-

tober 2006. ½Ê

½)" Jonasson, M. and Wallmark, O., ‘Control of electric vehicles with autonomous corner modules: implementation aspects and fault han-

dling’, submitted for publication, 2006. PaperC Æ`h ÆW² Jonasson, M. and Andreasson, J., ‘Exploiting autonomous corner modules to resolve force constraints in the tyre contact patch’, sub-

mitted for publication, 2007. PaperD